58#ifndef INCLUDED_volk_16ic_convert_32fc_a_H
59#define INCLUDED_volk_16ic_convert_32fc_a_H
66static inline void volk_16ic_convert_32fc_a_avx2(
lv_32fc_t* outputVector,
68 unsigned int num_points)
70 const unsigned int avx_iters = num_points / 8;
71 unsigned int number = 0;
72 const int16_t* complexVectorPtr = (int16_t*)inputVector;
73 float* outputVectorPtr = (
float*)outputVector;
78 for (number = 0; number < avx_iters; number++) {
79 cplxValue = _mm_load_si128((__m128i*)complexVectorPtr);
80 complexVectorPtr += 8;
82 outValInt = _mm256_cvtepi16_epi32(cplxValue);
83 outVal = _mm256_cvtepi32_ps(outValInt);
84 _mm256_store_ps((
float*)outputVectorPtr, outVal);
89 number = avx_iters * 8;
90 for (; number < num_points * 2; number++) {
91 *outputVectorPtr++ = (float)*complexVectorPtr++;
101 unsigned int num_points)
104 for (
i = 0;
i < num_points;
i++) {
114#include <emmintrin.h>
118 unsigned int num_points)
120 const unsigned int sse_iters = num_points / 2;
127 for (number = 0; number < sse_iters; number++) {
134 _mm_store_ps((
float*)_out, a);
138 if (num_points & 1) {
147#include <immintrin.h>
151 unsigned int num_points)
153 const unsigned int sse_iters = num_points / 4;
158 unsigned int i, number;
160 for (number = 0; number < sse_iters; number++) {
171 _mm256_store_ps((
float*)_out, a);
176 for (
i = 0;
i < (num_points % 4); ++
i) {
190 unsigned int num_points)
192 const unsigned int sse_iters = num_points / 2;
200 unsigned int i, number;
202 for (number = 0; number < sse_iters; number++) {
203 a16x4 = vld1_s16((
const int16_t*)_in);
205 a32x4 = vmovl_s16(a16x4);
206 f32x4 = vcvtq_f32_s32(a32x4);
207 vst1q_f32((float32_t*)_out, f32x4);
211 for (
i = 0;
i < (num_points % 2); ++
i) {
220#ifndef INCLUDED_volk_16ic_convert_32fc_u_H
221#define INCLUDED_volk_16ic_convert_32fc_u_H
227#include <immintrin.h>
229static inline void volk_16ic_convert_32fc_u_avx2(
lv_32fc_t* outputVector,
231 unsigned int num_points)
233 const unsigned int avx_iters = num_points / 8;
234 unsigned int number = 0;
235 const int16_t* complexVectorPtr = (int16_t*)inputVector;
236 float* outputVectorPtr = (
float*)outputVector;
241 for (number = 0; number < avx_iters; number++) {
242 cplxValue = _mm_loadu_si128((__m128i*)complexVectorPtr);
243 complexVectorPtr += 8;
245 outValInt = _mm256_cvtepi16_epi32(cplxValue);
246 outVal = _mm256_cvtepi32_ps(outValInt);
247 _mm256_storeu_ps((
float*)outputVectorPtr, outVal);
249 outputVectorPtr += 8;
252 number = avx_iters * 8;
253 for (; number < num_points * 2; number++) {
254 *outputVectorPtr++ = (float)*complexVectorPtr++;
261#include <emmintrin.h>
265 unsigned int num_points)
267 const unsigned int sse_iters = num_points / 2;
274 for (number = 0; number < sse_iters; number++) {
281 _mm_storeu_ps((
float*)_out, a);
285 if (num_points & 1) {
295#include <immintrin.h>
299 unsigned int num_points)
301 const unsigned int sse_iters = num_points / 4;
306 unsigned int i, number;
308 for (number = 0; number < sse_iters; number++) {
319 _mm256_storeu_ps((
float*)_out, a);
324 for (
i = 0;
i < (num_points % 4); ++
i) {
static void volk_16ic_convert_32fc_generic(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:99
static void volk_16ic_convert_32fc_u_avx(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:297
static void volk_16ic_convert_32fc_a_avx(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:149
static void volk_16ic_convert_32fc_u_sse2(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:263
static void volk_16ic_convert_32fc_neon(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:188
static void volk_16ic_convert_32fc_a_sse2(lv_32fc_t *outputVector, const lv_16sc_t *inputVector, unsigned int num_points)
Definition: volk_16ic_convert_32fc.h:116
#define __VOLK_PREFETCH(addr)
Definition: volk_common.h:62
#define lv_cimag(x)
Definition: volk_complex.h:89
#define lv_cmake(r, i)
Definition: volk_complex.h:68
#define lv_creal(x)
Definition: volk_complex.h:87
float complex lv_32fc_t
Definition: volk_complex.h:65
short complex lv_16sc_t
Definition: volk_complex.h:62
for i
Definition: volk_config_fixed.tmpl.h:25